Page 119 - untitled

юॆࠊ

Wu, Jan-Jan ޼Ӻᔊʧ Research Description
Research Description
޼Ӻᔊʧ

Ңٙ˴ࠅ޼Ӻጳሳ̻݊Бၾʱ౳όࠇၑf˴ࠅ My major research interests include high-per- output techniques can help solve this problem by
formance cluster computing and Grid computing. creating multiple data paths between memory and
޼Ӻሙᕚܼ̍˸ࡈɛཥ໘ᓉණމ˴ʘᓉණό৷஺ࠇ disks. In this work we have developed optimization
My current research in high-performance techniques for I/O resource assignment and paral-
ၑ˸ʿၣࣸࠇၑf cluster computing focuses on optimization of col- lel I/O scheduling, including an optimal I/O server
lective communication and parallel I/O. Collec- placement strategy for switch-based clusters, and an
tive communication, such as multicast and barrier optimal scheduling algorithm for parallel I/O with
ίᓉණό৷஺ࠇၑ˙ࠦd޼Ӻࠠᓃܼ̍ίᓉණ
synchronization, provides important functionality replicated data.
όၣ༩ᐑྤʕʘ༟ࣘෂ፩௰Գʷe˸ʿ˸̻Бʷ˙ for many applications. We have developed a num-
ber of algorithms for collective communication on In Grid computing research, we address the Research Fellows
ό౤؁Ꮶࣩ༟ࣘᛘᄳࣖଟ(̻БI/O)fމ౤؁Ꮠ͜ழ cluster systems, including multicast algorithms for problems of data replica placement in Data Grids
heterogeneous cluster systems and large-scale clus- given the traffic pattern and locality requirements.
᜗ί໊ଡ଼ᐑྤʕʘੂБࣖଟd༟ࣘෂ፩ʘ௰Գʷ݊ ter systems, an agent-based scheduling strategy for We propose a placement algorithm that finds the op-
optimizing multicast performance on switch-based timal locations for the replicas so that the workload
ɓ฽ࠠࠅٙሙᕚf০࿁ϤਪᕚҢࡁ౤̈ɓଡ଼Ҟ஺˲
networks, and a new, simple network topology, for among these replicas is balanced. We also propose
৷ࣖଟစၑجd̙͜˸ྼତε၇ʔΝ೷᜗ᐑྤʕʘ supporting deadlock-free, contention-free adaptive a new algorithm to decide the minimum number of
routing on clusters. replicas required when the maximum workload ca-
Research Fellows
͉
ਿ
༟
ࣘ
ਿ ͉ ༟ ࣘ Ҟ஺༟ࣘෂ፩fҢࡁɰ౤̈ɓ၇ቇ͜׵ܔໄᓉණό Due to the performance gap between CPU pacity of each replica server is known. In addition,
ၣ༩ٙၣ༩ஹഐןዎfϤஹഐןዎՈԄλʘᅰኪ׌ and disk read/write, the performance of carefully we propose a new model in which the request must
be served with quality-of-service guarantee. In this
tuned parallel programs can slow down dramatically
ᔖcc၈j ਓ޼Ӻࡰ ሯ˲฽׸ႡЪf new model we also propose algorithms that ensure
when they read or write files. Parallel input and both workload balancing and quality of service.
Associate Research Fellow (2000--)
௰৷ኪዝj Ph.D., Computer Science, Ϥ̮dίତʦٙ೷᜗ҦஔࠢՓʘɨdှဿᛘᄳ
Yale University (1995) ٙ஺ܓჃˢʕ̯ஈଣኜٙ༶ၑ஺ܓࠅໝܝ੻εdν
ཥcc༑j+886-2-2788-3799 ext. 1610 ؈೻ό඲ਂ᎖ᐿٙ፩̈ɝਗЪdှဿᛘᄳٙࣛගֻ
Selected Publications
ෂccॆj+886-2-2782-4814 ֻϓމ዆ࡈ೻όੂБࣖଟٙଧ᎕fνО౤؁I/Oٙ஺ Selected Publications
ཥɿڦᇌjwuj@iis.sinica.edu.tw ܓ݊Սʔ࢙ᇠٙሙᕚf޴Ч׵̻БࠇၑdI/Oਪᕚɰ 1. P. Liu, Jan.-Jan. Wu, Y.-F. Lin and S.-H. Yeh, A simple incremental 11. P. Liu, S.-K. Wang, Jan-Jan Wu, and Y.-M. Zhung, A Grid-based game
network topology for switched-based networks, Internationl Parallel tree evaluation system, Proceedings of the First Workshop on Grid
ၣccࠫjhttp://www.iis.sinica.edu.tw/pages/wuj ̙ᔟп׵̻БʷҦஔ(̻БI/O)fҢࡁί̻БI/Oٙ޼ and Distributed Processing Symposium (IPDPS), Cancun, Mexico, Technologies and Applications, HsinChu, Taiwan, Dec. 2004.
May 2001. 12. M.-B. Chen, T.-R. Chuang, and Jan-Jan Wu, Parallel Divide-and-Con-
ӺͦۃʊϞɨΐٙϓ؈j€ɓᔟ͟ʱৣ፩̈ɝ༟ 2. P.-Z. Lee, Jan-Jan Wu, and C.-H. Chang, Parallel implicit Euler solver quer Scheme for 2D Delaunay Triangulation, accepted by Concurency
on homogeneous and heterogeneous computing evnironments, 15th and Computation: Practice and Experience, June 2005.
๕ʘ௰Գʷ˸ಯˇ༟ࣘෂ፩ٙඎၾࣛගd(ɚ)ਗ਼༟ AIAA Computational Fluid Dynamics Conference, Anaheim, CA, June 13. Y.-F. Lin, C.-M. Wang, and Jan-Jan Wu, Optimizing I/O Server Place-
• Assistant Research Fellow, Academia Sinica 2001. ment for Parallel I/O on Switch-Based Irregular Networks, accepted
(1995--2000), ࣘቇʲʲ௲ʱ౳Ց΢ࡈᎷπண௪d˸ಯˇჃ൷π՟ 3. P.-Z. Lee, Jan-Jan Wu, and C.-H. Chang, Partitioning unstructured by The Journal of Supercomputing, April 2005.
meshes for homogeneous and heterogeneous parallel computing en- 14. P. Lee, C.-M. Wang, and Jan-Jan Wu, Compiler and Runtime Paral-
• M.S., Computer Science, Yale University (1991), ٙϣᅰd(ɧ)ቇ຅ર೻೷ဿ༟ࣘᛘᄳʘනҏ˸ഛ͜ vironments, International Conference on Parallel Processing (ICPP), lelization Techniques for Scientific Computations on Distributed
Vancouver, British Columbia, August, 2002. Memory Parallel Computers, Chapter 8, High-Performance Comput-
• B.S., Computer Science and Information ၣ༩ձ೷ဿٙ᎖ᄱd˸ᐵ೵̻БI/OٙੂБࣛගf 4. P. Liu, Jan-Jan Wu, and T.-S. Yang, Locality-Preserving Dynamic ing: Paradigm and Infrastructure, John Wiley and Sons, edited by
Load Balancing for Data-Parallel Applications, the Special Issue of Laurence T. Yang and MinYi Guo, 2005.
Engineering, National Taiwan University (1985). Parallel and Distributed Systems, Journal of Information Science and 15. Jan-Jan Wu, Yih-Fang Lin, Efficient Distributed Algorithms for Paral-

ίၣࣸࠇၑ˙ࠦdҢࡁઞীίData Gridӻ୕ Engineering, Vol. 18, pp. 1037-1048, 2002 lel I/O Scheduling, IEEE International Conference on Parallel and
5. Y.-F. Lin, P. Liu, and Jan-Jan Wu, Efficient agent-based multicast on Distributed Systems, Japan, July, 2005,

ʕٙ༟ࣘልႡၾʱৣਪᕚfͦۃʊ೯̙࢝ᆽڭ༟ࣘ wormhole switch-based irregular networks, International Parallel and 16. G-S Hsu, P. Liu and Jan-Jan Wu, I/O Processor Allocation for Mesh
Distributed Processing Symposium (IPDPS), April 2003. Cluster Computers, IEEE International Conference on Parallel and
6. Jan-Jan Wu, D.-W. Wang, Y.-F. Lin, Placement of I/O servers to im- Distributed Systems, Japan, July, 2005,
У؂ኜʈЪࠋ༱̻ፅٙልႡၾʱৣ˙جd˸ʿ̙ί
prove parallel I/O performance on switch-based clusters, International 17. Jan-Jan Wu and P. Liu, Distributed scheduling of parallel I/O in the
Conference on Supercomputing, San Francisco, June, 2003. presence of data replication, ACM/IEEE International Parallel and
ഗ֛༟ࣘУ؂ኜঐ࢙஢௰ɽʈЪࠋ༱ඎٙૢ΁ɨӔ 7. P. Liu, D.-W. Wang, Jan-Jan Wu, Efficient parallel I/O scheduling in Distributed Processing Symposium (IPDPS), April, 2005.

the presence of data duplication, International Conference on Parallel 18. P. Liu,Y.-F. Lin and Jan-Jan Wu, Optimal Placement of Replicas in
֛௰ʃٙ༟ࣘልႡ΅ᅰf Data Grid Environments with Locality Assurance, Proceedings of the
Processing (ICPP), Oct. 2003.
8. Jan-Jan Wu, S.-H. Yeh, and P. Liu, Efficient Multiple Multicast on Second Workshop on Grid Technologies and Applications, Taiwan,
Heterogeneous Network of Workstations, Journal of Supercomputing, Dec. 2005.
Vol. 29, pp. 59-88, 2004. 19. Y-F. Lin , Z-H Kang, P. Liu and Jan-Jan Wu, An Optimal Scheduling
9. M.-B. Chen, T.-R. Chuang, and Jan-Jan Wu, Efficient Parallel Imple- Algorithm for an Agent-Based Multicast Strategy on Irregular Net-

mentations of Near-Delaunay Triangulation with High Performance works, International Conference on Grid and Pervasive Computing
Fortran, Concurency and Computation: Practice and Experience, Vol. (GPC), Taiwan, May 2006.
16, No. 6, June 2004. 20. P. Liu and Jan-Jan Wu, Optimal Replica Placement Strategy for Hi-
10. Y.-F. Lin, C-.M. Wang, and Jan-Jan Wu, Optimizing Server Placement erarchical Data Grid Systems, International Conference on Cluster
for Parallel I/O on Switch-Based Irregular Networks, ACM/IEEE Computing and the Grid (CCGrid), Singapore, May, 2006.
International Symposium on Parallel and Distributed Processing and
Applications, LNCS 3358, Hong Kong, Dec. 2004.

108 109